| Measurement technology is one of the foundation of modern industrial technol-ogy. Measurement precision and efficiency play an important role on the development of manufacturing industry. With the improvement of industrial manufacturing technol-ogy and processing technology, the requirements of measuring precision and speed are more and more strict. In recent years, the development trend of measuring technology mainly reflected in the following directions:non-contact and online measurement, high precision, intelligence and so on. With the rapid development of machine vision, a new detection technology named visual measuring technique is formed, which applied the computer vision technology to measurement technology.Vision measurement technology is based on machine vision technology, which utilizes information in images to measure the displacement or acceleration. This tech-nology is a blend of optoelectronics, computer technology and image processing tech-nology and other modern science and technology. And it formed the opto-mechatronics integration vision measurement system with the characteristics of high efficiency, high precision, non-contact, long distance measurement. It is a big supplement to traditional measurement methods and it has been widely used in industry, military, medical and other fields. This paper takes the vision measurement system as the research object, and establishes the high-speed video acquisition device through hardware selection, and then studied video image analysis techniques and fast algorithms for displacement ex-traction.First, this paper gave an analysis of the hardware selection, and constructed the visual measurement hardware system based on high-speed camera. This system can be used to quickly capture the high-speed video. For the captured vibration video, the core of the visual measuring system is to extract vibration information from the video rapidly. Conventional means of image processing is commonly used in image analysis, which can complete various scenarios motion image analysis and video information retrieval through flexible combination of a variety of image processing algorithms. Subsequent steel cantilever on dynamic experiment shows the effect of conventional image pro-cessing method in practical application. However, cantilever beam experiments also exposes many problems of the conventional image processing techniques, such as pa-rameters are difficult to adjust, the whole procedure is tedious and error-prone, and lighting and noise has a direct impact on the performance of algorithm.To solve the problems of conventional image processing, this paper studied from template matching, data dimension reduction, mathematical analysis and image align- ment. Then a variety of simple and effective extraction algorithms were proposed, for example, the extraction algorithm based on NCC, SVD decomposition, Taylor expan-sion and Lucas-Kanade alignment. These algorithms can extract vibration information from video without complex image processing procedures, and they can quickly com-plete the extraction of the vibration displacement by simply selecting tracking area. These algorithms have the characteristics of easy to implement, strong commonality and high efficiency, and they can be employed automatically without too much param-eters to set.The NCC algorithm calculated the correlation coefficient matrix between the tem-plate image and the original image, and the index of the maximum value of the cor-relation coefficient matrix corresponds to the real position of the moving object. This algorithm is simple, intuitive and reliable. The SVD algorithm belongs to the dimension reduction algorithm. The dimension of image information can be reduced significantly through the singular value decomposition and k component signals are obtained even-tually, which reveals the hidden structural information behind the vibration video. This algorithm can extract displacement signal fast and automatically, the only requirement is to set the analysis area before processing. The Taylor expansion theory explains the physical meaning of the SVD extraction algorithm, and then a high precision sub-pixel extraction algorithm based on Taylor is proposed. The algorithm avoids the singular value decomposition and the image basis can be obtained directly by difference oper-ator, which make a further simplify algorithm process. The Lucas-Kanade algorithm introduced the concept of image alignment to extract vibration, and a more general extraction algorithm is proposed, which not only can realize the extraction of displace-ment, but also can complete rotation and deformation calculation. Finally, this paper made an integration between the Taylor algorithm and the Lucas-Kanade algorithm and proposed an modified reverse combination algorithm, which further enhances the ex-traction efficiency and makes the extraction time of each frame down to below 0.1 ms. This provides a possible method to real-time analysis of more than 1000Hz high-speed video.Based on the efficient displacement extraction algorithms, this paper developed vi-sual measurement software on the computer platform and mobile platform. By using the software a lot of experimental research are made. First, an experiment on high-precision displacement platform is carried out to verify the accuracy of the visual measurement system, and high accuracy results are achieved. Many application researches in the lab-oratory environment and outdoor environment are then carried out. Chair vibration and fan speed experiment verified the effect of mobile visual measurement software, and motor experiment verified that the mobile phones can accomplish the function of the vibration measurement without target. Outdoor sound barriers, Nanfeihe river bridge and forklift steering wheel vibration experiments showed the high-speed vision mea-surement software on computer can realize distant measurement. These experiments implied that the visual measurement is a reliable, effective way of measuring, and has important practical significance.
|
PDF Full Text Request